Consumer AR apps (built on ARKit, ARCore or WebXR) anchor 3D overlays to the real world in one of two ways: marker-based tracking locks onto a printed reference card placed near the hive, while markerless tracking uses the camera feed plus the device's own motion sensors to place content directly onto recognised features. Marker-based is more reliably accurate but needs physical setup in advance; markerless is more convenient but drifts more in the variable light and visual clutter typical of an apiary.
confidence = base(mode) − distPenalty − lightPenalty
overlay error (mm) ≈ 3 + (100 − confidence) × 1.4
- Tracking mode — marker-based starts near 96% confidence and barely reacts to light; markerless starts near 74% and is much more sensitive to both distance and glare.
- Device distance — every tracking method has a sweet spot; moving too close or too far from the hive raises the distance penalty and the position error.
- Outdoor light — deep shade and harsh midday glare both push the overlay off target, markerless far more than marker-based.
- Focus target — panning from the hive frame to a forage plant simulates the app re-detecting a different real-world subject, exactly like walking from an inspection to check what the colony is foraging on.
Real-world relevance: this confidence/error trade-off is why field-tested beekeeping AR apps default to marker-based tracking for precise tasks like locating a queen cell, and fall back to markerless only for looser tasks like general plant identification.